Skip to content
XALANTEdge

Xalant Edge · reference campus

Edge

Compute where demand and power live.

High-density, modular AI and Web3 compute deployed at the edge — close to where demand and baseload power live.

You bring the use case. Xalant is being built to deliver the site, the power, the backhaul and the compute as one platform.

Power · off-grid by designComms · satellite backhaul by design

Reference render of an Edge in a Box container on the Xalant campus, seen from the access road.

reference render

Edge district · camera stop 01
  • IT white space
  • N+1 precision air
  • Meet-me room and uplink
Plan view of the Edge in a Box container: the rack line, the cooling plant, the UPS and battery room, the switchgear tie-in and the meet-me room.

reference render

Edge in a Box · plan viewreference plan, not a photograph

The pod

What ships in every container

Every element ships as a factory-integrated container — site work is pads, conduit and connections, not construction.

  1. IT white space

    The rack line, with hot and cold aisle containment. This is the part the whole enclosure is sized around.

  2. N+1 precision air

    Redundant cooling, specified for free cooling on ambient air below the design threshold and pre-piped for direct-to-chip liquid.

  3. UPS and battery room

    Ride-through and dual-feed distribution, so a transfer between sources is not an outage. Early smoke detection and clean-agent suppression sit here too.

  4. Switchgear and genset tie-in

    Where on-site generation lands. Gensets, storage and solar are designed onto one bus rather than three.

  5. Meet-me room and uplink

    Where the satellite path meets the terrestrial network, and where fibre lands in a later phase. A site is designed to be reachable before any trench is dug.

  6. Resilience

    Access control · CCTV · clean-agent suppression

Deployment

Four phases to operational status.

The first build proves the kit. Every site after it repeats it.

  1. Identify

    Site parameters, power, latency and land in under-served US markets. A parcel either clears the published gates or it does not, and that is settled before anyone spends money on it.

    Duration varies by market and is published per market once the first one is done.
  2. Land and license

    Acquire the parcel, then zoning, environmental and building permits. This is the phase the jurisdiction controls, not Xalant.

    Local approval time sits outside every duration quoted on this site.
  3. Build and cool

    Pads, conduit and connections, then the factory-integrated modules land on them. Cooling is commissioned with the rack line rather than after it.

  4. Onboard and hand over

    Commissioning, acceptance against the design targets, then the workload moves in. AI training, inference, HPC and Web3 nodes are all planned to be carried on the same kit.

NoteDurations exclude local approval time and assume no grid interconnection.

Workloads

Purpose-built for AI training and inference, HPC and Web3 processing.

Sited at scale or close to demand, removing reliance on the distant core.

AI training

The densest rack line Xalant designs. The interconnect never leaves the hall, so what a training run needs from the site is power and cooling rather than backhaul.

Cooling · Direct-to-chip liquid
Inference

The opposite shape. Every served response crosses the uplink, so an inference site is sized by the backhaul as much as by the load — and by how close it sits to the demand.

Cooling · N+1 precision air
HPC

Bulk in, results out, in scheduled windows. Batch work tolerates distance from the core better than almost anything else, which is what makes an under-served market a good place to put it.

Cooling · N+1 precision air
Per-rack density, per-rack compute and the mature-site GPU count are design targets held in `content/stats.json` and are published when their basis and their as-of date clear.

Locations

Four under-served US markets.

One proof-of-concept build, three planned.

  • South OregonPOC build-out · Q1 2027
  • MilwaukeePlanned
  • PennsylvaniaPlanned
  • North GeorgiaPlanned

LedgerDeployed · contracted · pipeline, with an as-of date, on the map page

Estimator

Size a site in three inputs.

IT load, workload and market.

The tool returns pods, electrical megawatts, gateway pads and a time-to-power tier, and it shows every line of the arithmetic. Every constant it uses is published beside the answer with its source and its date.

It sizes a site. It quotes nothing — pricing is not settled and a number here would be invented.

News

From the Edge newsroom

No Edge announcements yet — see all Xalant news →

See all Xalant news
Line drawing of the Xalant campus: the hall row, the substation and the gateway dish, derived from a reference render.

Reference render

The campus, drawnAI render — reference only

Site & capacity · Request

Bring us a use case. We bring the site, the power, the backhaul and the compute as one contract.